We have performed CDKN2A mutation analysis on subjects from melanoma families from the NCI Family Cancer Registry studied by Peggy Tucker and Alisa Goldstein of the Genetic Epidemiology Branch (GEB) for many years. We recently identified a novel mutation within the amino acid coding portion of CDKN2A (D153cspm) that alters splicing in the same manner as an earlier mutation we identified (IVS2+1g>t). Within the NCI families, three pedigrees, with many cases of melanoma and some of the highest lod scores ever observed for linkage to the p16 locus on chromosome 9, do not have mutations within the mRNA encoding portion of CDKN2A, nor in the more than two kilobases of upstream potential regulatory region sequences based on analysis in multiple laboratories, including ours. While it is possible that these two linked but mutation negative families are due to another gene in this region, a more likely explanation is that they contain a mutation that is not detectable with the PCR-based screening techniques we are using. We have used a number of different primer pairs and sequenced both genomic DNA and cDNA, so it is unlikely that there are undetected mutations in the mRNA encoding portions of the gene. These linked but mutation negative families are extremely unlikely to have larger genomic deletions in the region because they carry the CDKN2A amino acid coding polymorphism A148T that segregates with disease in the families, and the affected individuals are heterozygous for microsatellite markers flanking the locus. They may have an intronic mutation that affects splicing but is not included when sequencing the exons, similar to the deep intron splicing mutation recently identified in a number of melanoma families, particularly from Great Britain. We have analyzed our samples directly for this intronic mutation, and have performed RT-PCR amplifying the entire coding portion of the gene, and do not observe any variant products that would represent splicing differences. There are no cells available from members of these families that express p16 (measurable by Western blot), such as fibroblasts or melanoma cell lines, and Westerns performed on lymphocyte lysates have shown no detectable p16 protein. We are continuing to study these three families and to collaborate with GEB investigators conducting melanoma case-control studies. We recently identified a p14ARF splice site mutation (intron 1-beta +1G>C) as the likely causative mutation in one family, but the other two families are still being evaluated.